AI SDR vs Human SDR: What a Meeting Actually Costs in 2026

A loaded US SDR costs $110,000 to $130,000 in year one. An AI SDR stack costs $1,400 to $1,750 a month in cash. And the category those AI tools belong to churns at 50–70% a year — roughly twice the annual attrition of a human SDR team, and about ten times a normal SaaS product. The CEO of the company that coined the term "AI SDR" said in 2026 that AI SDRs do not work in their current form.

Published 9 August 2026 · Last updated 9 August 2026 · Compensation and performance data from The Bridge Group's 10th edition (6 February 2025, 351 B2B companies) and RepVue (2 August 2026, 8,422 submissions). Every modelling assumption is labelled where it appears.

Disclosure first, because it should change how you read everything below. We build Lead Scorer. Lead Scorer is a product in the category this article criticises. Writing about a market's churn rate, its opaque annual contracts and its unverifiable claims while quietly exempting yourself is the same move the underlying research flags as the defining problem with the literature on this subject: almost every number comparing AI to human sales development comes from a company selling the AI. That includes us. So the rule for this piece is that our product appears only in the two places the data puts it — nowhere else, and with no superlative attached.

What the category's own numbers say

Start with the failure data, because it is the part vendors do not publish and the part that changes the arithmetic.

• Category churn: 50–70% a year (UserGems). A human SDR team's median annual attrition is 40% (The Bridge Group 2025, split 13% involuntary, 11% voluntary, 16% promotion). The tool that replaces the rep leaves faster than the rep.
• Around 2% of companies reach a durable AI SDR implementation, per the GTM AI Podcast.
• 42% of companies abandoned most of their AI initiatives in 2025, up from 17% in 2024 — S&P Global Market Intelligence, survey of over 1,000 companies. This is an observation.
• Gartner forecasts that over 40% of agentic AI projects will be cancelled by end of 2027 (press release 25 June 2025, from a January 2025 survey of 3,412 respondents), citing cost, unclear business value and inadequate risk controls. This is a forecast, not a result. The two numbers sit close enough together — 42% and 40% — that they get merged constantly. One measures what happened last year, the other projects what will happen by 2027.

Then there is the case everyone in the category knows about. A TechCrunch investigation published on 24 March 2025 reported that 11x had displayed ZoomInfo and Airtable as customer logos when neither was a customer, that stated ARR of roughly $10–14M compared with around $3M surviving the three-month break clause, and that former employees described churn of 70–80%. 11x itself claims 79% retention. Founder Hasan Sukkar was replaced by CTO Prabhav Jain in May 2025.

11x is still operating in 2026 — around 86 employees per Tracxn, roughly $76M raised in total, no new round since the $50M Series B led by a16z in November 2024. The most quotable line about the category in 2026 came from its own chief executive:

I don't think AI SDRs work in their current form, and you're hearing that from the CEO of the company that invented the word.
— Prabhav Jain, CEO of 11x, 2026

One clarification, stated plainly because the internet gets it wrong. No lawsuit by ZoomInfo against 11x has been confirmed as filed. The widely circulated "ZoomInfo lawsuit" of 2026 is a securities class action — Tejeda v. ZoomInfo, W.D. Wash., June 2026 — which has nothing to do with 11x. The logo dispute has never been confirmed as having become litigation.

What a human SDR actually costs

The number a founder usually quotes is the OTE, and the OTE is the number the rep sees, not the number the company pays.

Geography	Base	OTE	Source (date)
US, national median	$60,000	$85,000	RepVue, 2 August 2026 (8,422 submissions)
US (The Bridge Group)	$55,000	$80,000	Bridge Group 10th ed., 6 February 2025 (351 B2B companies)
US, San Francisco	$60,000	$90,000	RepVue SF, 24 March 2026
US, top performers	—	$129,000 – $131,581	RepVue, 2026
UK, national	£39,523 – £40,269	£60,757	RepVue GB, 8 January 2026
UK, London	£38,000 – £45,000	£50,000 – £70,000	Live Digital, 2025–26
France, average package	€26k – €35k	€45k – €55k	Uptoo, 2026
France, median OTE	€35,500 (25th pct)	€45,748	Glassdoor FR, December 2025

A detail worth noticing in that table: US geography has stopped mattering much. San Francisco and New York now pay 5–15% more than the rest of the country, against 20–30% before 2023. Remote work compressed the premium.

Everything that gets added on top of the OTE:

• Employer multiplier: 1.25× to 1.4× on base pay (the MIT/Hadzima rule). A $60,000 base becomes $75,000–$84,000 before a single tool is bought. US federal data from September 2025 puts benefits at about 42 cents per salary dollar.
• Tools: $371 per SDR per month on average — The Bridge Group, via QuotaPath. That is $4,452 a year, and it is before ZoomInfo, which starts around $15,000 a year and runs past $60,000.
• Management: roughly $22,800 a year per SDR — and this one is our calculation, not a published number. SDR manager OTE is $146,000 and the ratio is 6.4 SDRs per manager, both from The Bridge Group 2025. Dividing one by the other is us, not them. The ratio has been falling, from 8 in 2021–2023.
• Hiring: about $4,700 per hire (SHRM), plus equipment and onboarding.
• Ramp: 3.0 months at roughly 50% productivity, so about 1.5 months of lost output in year one.

Total year one: $110,000 to $130,000 in the US, or roughly €65,000–€85,000 loaded in France plus tools and management. That works out to $9,200–$10,800 a month averaged across the first year.

Two source conflicts we are not going to hide inside that total, because both move it. Ramp time is reported at 3.0 months by The Bridge Group's 2025 edition, 3.2 months in the 2024 edition, and 3.9 to 5.5 months by assorted third parties. We use 3.0 because it is the most recent primary source reading its own survey; at 5.5 months the year-one production loss doubles. Quota attainment splits by who was asked: 60% when companies are surveyed (The Bridge Group), 53–57% when reps are surveyed directly (RepVue, and Everstage/RepVue's Cloud Sales Index Q4 2024 at 53.2%). We are not picking one. Company surveys describe the team as budgeted, crowdsourced rep data describes what individuals live; carry the full 53–60% band.

For reference on what that spend buys: median pipeline of $3.78M per SDR per year (25th–75th percentile $1.9M–$6.4M), 112 activities a day (44 calls, 41 emails, 19 LinkedIn, 8 SMS) producing 4.1 quality conversations, and a median monthly Stage 0 quota of 10 meetings.

What an AI SDR costs, including the parts not on the pricing page

Tool	Price	Model
AiSDR	$250 (Solo) / $900 (Explore) / $2,500 (Scale) per month	Seat + $0.75/message, billed quarterly
Artisan (Ava)	~$1,000 – $2,500/month (reported, not published)	Usage, no platform fee
11x (Alice)	$36,000/year on its own page; $3,750/month billed annually per a competitor's writeup; ~$5,000/month reported by third parties	Per lead, annual
Qualified (Piper)	$40,000 – $68,000/year, plus a Salesforce stack at $30k–$60k (reported)	Annual "hire"
Regie.ai	$180/seat (10 minimum) or $499 (5 minimum)	Per seat
Salesforge (Agent Frank)	$499/month	Plan

Read the 11x row again. $36,000 a year and $3,750 a month billed annually cannot both be true, and both circulate with the vendor's name on them. Artisan has moved to quote-only pricing and third parties now report anything from $280 to $5,000 a month for it, a spread of about 18×. 11x, Artisan and Qualified are all sales-led: there is no firm price without a demo. Annual minimums are common — $36k at 11x, roughly $40k at Qualified, $35k at Regie — with two to four weeks of setup, sometimes billed.

Then the costs that sit underneath the subscription:

• Email infrastructure. The 2026 practice is 20–30 emails per mailbox per day, so 1,000 emails a day means 35–50 mailboxes across 12–25 domains. Pre-warmed mailboxes run about $4.99 each, standalone warmup $15–$29 per inbox per month, and a DIY setup of five domains and 10–15 mailboxes lands at $150–$320 a month. The per-mailbox daily limit is itself contested — 10–20, 20–30 and 40 all circulate, and it is worth noticing that the vendor recommending 40 sells mailboxes while the one recommending 10–20 sells warmup.
• Data. Apollo at $49/seat, Cleanlist at $79/month, ZoomInfo from $15,000 a year. A verified list gets twice the reply rate of an unverified one and five to six times that of a bought one (Cleanlist), and bounce rate is the single biggest separator between top and bottom performers.
• Deliverability risk. Around 47% of AI SDR deployments hit a domain reputation wall within 90 days and 21% never recover inbox placement (Smartlead/Instantly figures, reported via digitalapplied — vendor-sourced, so directional). A burned primary domain can take months to recover, sometimes never.
• Human supervision: 15–20 hours a week (SaaStr). Sequence review, reply handling, qualification, ICP tuning. This is the line that decides the comparison, and it almost never appears in a vendor's ROI calculator — it is the same weekly load as doing the outreach by hand.

The model: three scenarios, costed

Here are the assumptions, stated before the numbers rather than in a footnote, because two of them carry the result. Sourced: a 3.43% cold email reply rate (Instantly, 2026) for volume outreach, and a 5–8% band for targeted human outreach (Agentic Demand/Instantly). Assumed by us: 85% deliverability, positive replies at about 40% of all replies, reply-to-meeting at 40%, a minimal stack in each scenario, and founder time valued at $100/hour. That last one is doing a lot of work — halve it and the loaded cost of scenarios (a) and (b) halves while (c) does not move at all. These are orders of magnitude. They are not a quote.

(a) The founder does it himself

Apollo at $49, light email infrastructure at about $150, a sequencer at about $97 — roughly $300 a month. Fifteen to twenty hours a week, which at $100/hour is $6,500–$8,700 a month of opportunity cost. Realistic volume is 20–40 heavily personalised emails a day, so 600–1,200 a month. At 5–8% reply, 40% positive, 40% converting to a meeting, that is 0.8–1.3% of emails sent, or 6 to 10 meetings a month. Cash cost per meeting: $30–$50. Loaded: $700–$1,400.

(b) The founder plus an AI SDR stack

AiSDR Explore at $900 a month as a realistic quarterly-billed entry point, $300–$450 of email infrastructure, $200–$400 of data — $1,400 to $1,750 a month in cash. Plus the same 15–20 hours a week of supervision, which is another $6,500–$8,700 if you count it, taking the loaded figure to $8,000–$10,000. Volume of 1,000–1,500 emails a day is 25,000–30,000 a month; at 3.43% reply, 40% positive, 40% to meeting, roughly 21,000 delivered emails produce about 720 replies, 288 positive ones and 115 raw meetings. After filtering out false positives and duplicates, 45 to 115 qualified meetings a month. Cash cost per meeting: $12–$39, loaded (cash + time) $70–$220 — with lower downstream quality.

(c) The first human SDR hire

$9,200 to $10,800 a month averaged over year one, producing about 9 qualified meetings a month at steady state and roughly 7 a month averaged across year one once ramp is accounted for. Cash cost per meeting: $1,100–$1,500, with materially better meeting-to-opportunity conversion and show rate.

Scenario	Cash / month	Meetings / month	Cash per meeting	Loaded per meeting
(a) Founder, by hand	~$300 + time	6 – 10	$30 – $50	$700 – $1,400
(b) AI SDR stack	$1,400 – $1,750 + time	45 – 115	$12 – $39	~$70 – $220
(c) Human SDR	$9,200 – $10,800	~7 (year one)	$1,100 – $1,500	$1,100 – $1,500

On cash per meeting the AI wins by a large factor — roughly 30x to 90x depending where in each range you land. That is the number every vendor deck is built on, and it is real. What it does not price is that the meetings are not the same object. A vendor's own published case describes a client booking 47 meetings in six weeks that produced 4 opportunities, with the sender score falling from 95 to 72 (Prospeo). A controlled 90-day test reported by the GTM AI Podcast had AI-only outreach producing 847 meetings at 11% conversion to opportunity, while the hybrid setup generated 2.3× more revenue. If you want one metric to run this decision on, it is cost per opportunity, not cost per meeting.

The three thresholds that actually decide it

Threshold	Value	What it means
Addressable market	> 10,000 accounts	Below this, volume burns the list before it converts. At 200 touches a day and under 1% reply, a 2,000-buyer market is exhausted in two weeks.
ACV	< $25k · $25k–$50k · > $50k	Under $25k the AI can carry the first touch. Between $25k and $50k, hybrid — AI researches, human engages. Above $50k the human leads and the AI assists.
Sales cycle	Short vs complex	Short cycles tolerate automation. Complex ones put the value in discovery and objection handling, which is exactly where the automation has nothing to offer.

Where AI SDRs genuinely do not substitute for a person: discovery calls, objection handling, strategic and enterprise accounts, relationship building, closing, real qualification, and replying to positive replies — the conversation is the actual job. Also any niche market, where volume destroys the list faster than it converts it.

Where they are genuinely good: inbound speed-to-lead (seconds against a human's 42–47 hours), first-touch email volume, research and enrichment, and signal-triggered personalisation. That last claim comes from a vendor and should be weighted accordingly.

What we would actually tell a founder

1. Before product-market fit, do it yourself. Scenario (a), about $300 a month. The point is not volume, it is learning the market and the message. Do not pay $36,000 a year for an AI SDR before you know what converts. The exit signal is time: more than about 15 hours a week, sustained.
2. Large market, low ACV, short cycle: pilot something you can cancel. Quarterly billing over annual, a published price over a quote, contactable customer references. Measure qualified meetings and pipeline, not emails sent. Success threshold: cost per opportunity under $300–$500 and positive reply above 1%.
3. High ACV, complex sale, niche market: hire the human — or go nearshore / offshore at $1,500–$3,500 a month if budget is tight. Budget for three months of ramp at 50% and 40% annual attrition.
4. Hybrid is the best-supported configuration in the data. AI for first-touch volume and speed-to-lead, humans for replies, qualification and closing. UserGems puts 45% of teams there already, and it is the arm that won the controlled test.
5. Know your kill criteria before you sign. Cut the tool if a quarter passes with no qualified meetings. If reply rate drops below 1% or bounce goes above 3%, stop sending and clean the list before scaling. Never sign an opaque annual contract without a cancellable pilot.

Where this leaves us, since we sell in this category

The two places the data puts our product are the two places we will name it. Recommendation 2 favours tools that publish a price and let you leave. Lead Scorer meets that test in the narrow, checkable sense: the prices are on the pricing page, and there is a monthly plan, so no annual commitment is required to use it. There is also a discounted annual plan, and that one is a twelve-month commitment — which is the same structure we just spent this article warning about, minus the opacity. That is a description of our billing, not an argument that we work better. Recommendation 1 says the rational starting point before PMF is the founder doing outreach himself on a minimal stack, which is exactly what the model in this article shows — not what benefits a company selling an AI SDR subscription.

Nothing in this article establishes that Lead Scorer escapes the category's failure rate. The 50–70% churn figure is a category number, and we are in the category. If our own retention were the point of this article we would have to publish it, and this is not that article.

The numbers we refused to publish

This section exists because it is the most useful thing we can hand you. Four figures dominate every "AI SDR vs human SDR" page on the internet: a reply-rate comparison, a lead-to-meeting comparison, a revenue-per-rep comparison and a show-rate comparison. We went looking for their primary sources.

There are none. They trace back to vendor blogs with no stated methodology, no sample size and no date. We are not reproducing them even to debunk them, because a number printed in a debunk gets quoted out of the debunk. If you see a comparison table on this topic with suspiciously round, suspiciously favourable pairs of percentages, ask what survey produced them, how many respondents it had, and when. In our search, that question had no answer.

The other gaps we could not close, listed so you know what this article does not know: precise UK SDR OTE outside London; the exact French employer contribution rate for a given SDR salary after the RGDU reform that replaced the Fillon reduction on 1 January 2026 (it is degressive between 1 and 3 SMIC, so it has to be computed case by case — our 22–30% is a working assumption); and 11x's current retention and ARR, which have not been published since the TechCrunch investigation.

The companion piece prices the tools themselves rather than the labour: we priced 17 AI SDR tools for a team of one, and nothing exists under $99. If you want the case for doing the first hundred conversations by hand, the founders who sent fewer emails got more customers is the data behind scenario (a). And to test whether either motion can repay itself at your price point, the ACV-to-motion floor in our CAC payback breakdown turns a cost per seat into a minimum viable ARPA.